MLflow
Verified Oct 2, 2026 ActiveOpen-source platform for tracking ML experiments, registering models and tracing, evaluating and deploying LLM agents.
Listed on EffectHub ·
Pricing
- Pricing model
- 100% Free
The software is free under Apache 2.0 and can be self-hosted. Managed MLflow is sold by third parties such as Databricks, whose pricing is separate and was not checked here.
Use tool ↗About MLflow
MLflow began as an experiment-tracking and model-registry tool for classical machine learning and has grown into an engineering platform for LLM apps and agents. It records parameters, metrics and artifacts for each run, versions models and prompts, and captures OpenTelemetry-based traces of LLM calls. It adds evaluation with built-in metrics and LLM judges, an OpenAI-compatible AI Gateway for routing and cost control, and an Agent Server for hosting. The Apache 2.0 project works with Python, TypeScript, Java and R, runs on your own infrastructure, and is also offered as a managed service by cloud vendors such as Databricks.
- API
- Yes
- Last verified
- Oct 2, 2026
Verified sources
- Open source AI platform for agents, LLMs and models: tracing, evaluation, prompt registry, AI Gateway, experiment tracking; Apache 2.0mlflow.org ↗
- Repository created June 2018 with Apache-2.0 licencegithub.com ↗
Is MLflow legit?
We don’t label tools safe or unsafe. Below are the records we checked ourselves, each with the date we read it.
- Last checked
- Oct 2, 2026The site answered
- On the market since
- 2018The vendor's own material, read by us
- Operated by
- MLflow Project (LF Projects, LLC), United StatesOur research, card verified Oct 2, 2026
Pros & cons of MLflow
The default open-source choice for tracking experiments and registering models, now extended to cover tracing, evaluation and prompt management for LLM agents. It is free and runs anywhere, but it is infrastructure you operate rather than a polished SaaS. Teams that want a hosted experience should look at a managed offering or a commercial LLMOps vendor.
Pros
- Apache 2.0, self-hostable, no usage fees
- Covers classical ML, LLM tracing, evaluation, prompts and a gateway in one project
- Integrates with 100+ frameworks and OpenTelemetry
- Backed by a large community and the Linux Foundation project structure
Cons
- You run and scale the tracking server yourself
- Broad scope means a steeper learning curve than single-purpose tools
- Newer LLM features are still evolving quickly
Best for
- ML teams that need experiment tracking and a model registry without vendor lock-in
- Developers who want open-source tracing and evals for LLM apps and agents
- Organizations already on Databricks or other clouds that offer managed MLflow
Not for
- Teams that want a fully hosted tool with a polished UI and support out of the box
- People looking for a no-code workflow rather than a Python or Java SDK
Watch out: MLflow itself is free, but a production setup needs a database, artifact storage and access control that you must provide, and managed versions from cloud vendors are priced separately.